Surface passivation of crystalline silicon by dual layer amorphous silicon films
by
Dmitri S. Stepanov
A thesis submitted in conformity with the requirements for the degree of Master of Applied Science Graduate Department of Material Science and Engineering University of Toronto
c 2011 by Dmitri S. Stepanov Copyright
Abstract Surface passivation of crystalline silicon by dual layer amorphous silicon films Dmitri S. Stepanov Master of Applied Science Graduate Department of Material Science and Engineering University of Toronto 2011 The probability of recombination of photogenerated electron hole pairs in crystalline silicon is governed by the density of surface defect states and the density of charge carriers. Depositions of intrinsic hydrogenated amorphous silicon (a-Si:H) in dc saddle field (DCSF) PECVD system and hydrogenated amorphous silicon nitride (SiNx ) in rf PECVD system forms a dual layer stack on c-Si, which results in an excellent passivation of the surface and an anti-reflection coating. Response Surface Methodology is used in this work to optimize the deposition conditions of SiNx . Optimization of the response surface function yielded deposition conditions that materialized in a surface recombination velocity of less than 4cm/s. The BACH (Back Amorphous Crystalline silicon Heterojunction) cell concept makes use of this dual layer a-Si:H/SiNx stack to form a high efficiency photovoltaic device. The high quality passivating structure can result in the BACH solar cell device with more than 20% conversion efficiency.
ii
Acknowledgements I would like to express my deepest gratitude to all who made this work possible:
Prof. Nazir P. Kherani for his great support and guidance in this research; Dr. Honggang Liu for introducing the subject matter to me and the exposure to lifetime spectroscopy techniques; Barzin Bahardoust for the DCSF a-Si:H depositions; Keith Leong for the SE training and numerous useful discussions; Zahidur Chowdhury for the Sentaurus simulation of the BACH solar cell; Dr. Henry Lee, Dr. Edward Xu and Yimin Zhou for the introduction to the cleanroom fabrication techniques; Dr. Davit Yeghikyan and Dr. Tome Kosteski for the GEN 1 and GEN 2.1 DCSF PECVD training; Dr. A. Gougam for constructive discussions and reviews;
A special thank you is bestowed to all the members of the Advanced Photovoltaics and Devices group.
Finally, I would like to thank my family, and especially my parents, for their unfailing confidence in me.
iii
Contents
1 Introduction
1
1.1
Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1
1.2
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2
1.3
Research Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7
1.4
Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8
2 Mechanisms of Surface Passivation
10
2.1
Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10
2.2
Bulk Recombination . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
13
2.3
Surface Recombination . . . . . . . . . . . . . . . . . . . . . . . . . . . .
19
2.4
State of the Art Passivation Mechanisms . . . . . . . . . . . . . . . . . .
23
3 Surface passivation by multilayer amorphous films
28
3.1
Synthesis techniques for amorphous films . . . . . . . . . . . . . . . . . .
28
3.2
Characterization techniques . . . . . . . . . . . . . . . . . . . . . . . . .
41
3.3
Optimization of SiNx for a-Si:H/SiNx passivation . . . . . . . . . . . . .
50
4 Results
55
4.1
Chemical Passivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
55
4.2
Response surface plots . . . . . . . . . . . . . . . . . . . . . . . . . . . .
57
4.3
Annealing study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
69
iv
4.4
Spatial resolution of lifetime measurement . . . . . . . . . . . . . . . . .
5 Application to BACH solar cell
78 80
5.1
Back Amorphous-Crystalline silicon Heterojunction photovoltaic device .
81
5.2
Simulations of BACH solar cell . . . . . . . . . . . . . . . . . . . . . . .
81
6 Conclusions
84
6.1
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
84
6.2
Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
86
A Response Surface Methodology
87
A.1 Statistical experiment design . . . . . . . . . . . . . . . . . . . . . . . . .
89
A.2 Selecting response functions . . . . . . . . . . . . . . . . . . . . . . . . .
90
B Annealing study of multilayer amorphous silicon structures
95
Bibliography
99
v
List of Symbols Symbol
Description
Unit
AT L
fitting parameter in Tauc-Lorentz model
-
a-Si:H
hydrogenated amorphous silicon
-
B
radiative recombination coefficient
cm3 /s
BACH
back amorphous-crystalline silicon heterojunction
-
BSF
back surface field
-
Cn (Cp )
Auger recombination coefficient for electrons(holes)
cm6 /s
CT L
fitting parameter in Tauc-Lorentz model
-
c-Si
crystalline silicon
-
∆n
excess electron concentration
cm−3
∆
phase difference
-
D
carrier diffusion constant
cm2 s−1
Dp
hole diffusion constant
cm2 s−1
d
offset from surface, flat band conditions
µm
Dit
interface state density
cm−2
DCSF
dc saddle field
-
D0 ,D+ , D-
defect states in a-Si:H
-
η
energy conversion efficiency
%
E0T L
fitting parameter in Tauc-Lorentz model
-
Ec
conduction band
eV
Ef
Fermi energy
eV
Eg
band gap
eV
Et
trap energy
eV
EgT L
Tauc-Lorentz mathematical gap
eV
Ev
valence band
eV
vi
ehh
electron-hole-hole Auger process
-
eeh
electron-electron-hole Auger process
-
EDX
energy dispersive X-ray
-
inf
fitting parameter in Tauc-Lorentz model
-
f0
equilibrium Fermi distribution function
-
fabs
fraction of available photons
FZ
float zone wafer
-
Fe
iron
-
FSF
front surface field
-
I
intensity
W/m2
G
carrier generation rate
cm−3 s−1
JSC
short circuit current density
mA/cm2
k
Boltzmann’s constant
J/K
kr
extinction coefficient
-
K 0 ,K - , K +
defects in silicon nitride
-
λ
wavelength
nm
L
diffusion length
cm
MW-PCD
microwave-detected photoconductance decay
-
µn (µp )
electron (hole) mobilities
cm2 /(V s)
n
total electron concentration
cm−3
n0
initial electron concentration
cm−3
ni
effective intrinsic carrier concentration
cm−3
nr
refractive index
-
NA
acceptor concentration
cm−3
ND
dopant concentration
cm−3
Nt
density of defects
cm−3
n1
SRH recombination parameter
cm−3
vii
1sun Nphoton
number of available photons at 1 sun illumination
-
ONO
SiOx /SiNx /SiOx stack
-
p
total hole concentration
cm−3
p0
initial hole concentration
cm−3
p1
SRH recombination parameter
cm−3
p-a-Si:H
p-type hydrogenated amorphous silicon
-
PECVD
plasma enhanced chemical vapour deposition
-
Ψ
amplitude ratio
-
QSSPC
quasi steady state photoconductance
-
RCA
Radio Corporation of America
-
rp
normalized p-polarized component
-
rs
normalized s-polarized component
-
S
surface recombination velocity
cm/s
Seff
effective surface recombination velocity
cm/s
SC-1 (SC-2)
standard clean 1 (2)
-
SE
Spectroscopic ellipsometry
-
Si
silicon
-
SiO2
silicon dioxide
-
SiOx
plasma silicon oxide
-
SiON
silicon oxynitride
-
SiNx
hydrogenated amorphous silicon nitride
-
SRH
Shockley-Read-Hall
-
SRV
Surface Recombination Velocity
-
∆σ
excess photoconductance
Sm−1
σn
electron capture cross section
cm−2
σp
hole capture cross section
cm−2
T
Temperature
K
viii
TEM
Transmission Electron Microscope
-
TL
Tauc-Lorentz
-
τ
effective carrier lifetime
s
τAuger
Auger lifetime
s
τb
bulk carrier lifetime
s
τrad
radiative lifetime
s
τSRH
Shockley Read Hall lifetime
s
τn0
electron capture time constant
s
τp0
hole capture time constant
s
U
net rate of carrier recombination rate
cm−3 s−1
Uc
equivalent density of states at band edge
cm−3
UAuger
rate of Auger recombination
cm−3 s−1
Urad
rate of radiative recombination
cm−3 s−1
USRH
rate of Shockley Read Hall recombination
cm−3 s−1
VOC
open circuit voltage
mV
vth
thermal velocity
cm/s
W
wafer thickness
µm
xd
distance from the surface of wafer
µm
ix
List of Tables 3.1
DCSF Parameter settings - a-Si:H deposition parameters remained constant in this study . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2
Tauc-Lorentz Model Parameters for a-Si:H and SiNx - Taken from Jellison and Modine [27] . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3
36
49
D-Optimal Design for SiNx deposition - Summary of numeric factors used to generate experimental runs by Design-Expert software . . . . . .
54
4.1
Numerical results of experiments
58
4.2
Response surface model - Coefficient estimates and errors for τeff
4.3
Statistical metrics of the response surface model of minority car-
. . . . . . . . . . . . . . . . . . . . .
rier lifetime τeff . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4
59
60
Results of response surface optimization - predicted response values are confirmed experimentally . . . . . . . . . . . . . . . . . . . . . . . . .
62
4.5
Response surface model - Coefficient estimates and errors for nr . . .
64
4.6
Statistical metrics of the response surface model of index of refraction, nr
4.7
64
Response surface model - Coefficient estimates and errors for growth rate (nm/min)
4.8
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
66
Statistical metrics of the response surface model of SiNx growth rate, rgr
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
x
67
List of Figures 1.1
Solar modules installations - historical development of World cumulative PV power installed in main geographies, adapted from [28] . . . . . .
1.2
Costs of a typical c-Si solar cell - crystalline silicon amounts to 75% of costs at the cell level and 50% of solar module, [18] . . . . . . . . . . .
2.1
3
Intrinsic carrier lifetime in n-type silicon - ND = 1.53×1015 (3 Ωcm) and ND = 1.45 × 1014 (30 Ωcm) . . . . . . . . . . . . . . . . . . . . . . .
2.2
2
16
Electron energy band diagram showing four interaction mechanisms of free carriers with a defect level: electron emission(1), electron capture (2), hole capture (3), hole emission (4) . . . . . . . . . .
2.3
17
Bulk recombination mechanisms - Excess carrier density dependence of recombination mechanisms in the bulk of n-type silison ND = 1.53×1015 (3 Ωcm). SRH assumes defect levels of Fe in Si, with Nt = 1 × 1010 cm−3
2.4
18
a-Si:H/SiNx dual layer passivation - combines excellent interface passivation of a-Si:H with field-effect passivation and anti reflection properites of SiNx . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.5
a-Si:H/SiNx Dual layer passivation - summary of SRV values published in the literature . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1
26
27
Thermodynamic Model for surface defect formation - from Reddy et al. [45] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
31
3.2
Density of states in a-Si:H- adapted from [55] . . . . . . . . . . . . .
33
3.3
DCSF PECVD system - used for depositing a-Si:H films . . . . . . . .
34
3.4
Oxford Plasmalab 100 - rf PECVD system used for SiNx deposition .
39
3.5
QSSPC experimental setup - Block diagram of inductively-coupled photoconductance measurement technique and Sinton WCT-120 apparatus used to measure injection-level dependent carrier lifetime . . . . . . .
3.6
44
MW-PCD experimental setup - Block diagram of microwave reflected measurement technique and WT-2000 apparatus used to collect spatially resolved minority carrier lifetime measurements . . . . . . . . . . . . . .
3.7
45
Measurement principle of ellipsometry - The variation of light reflection with p- and s-polarizations is measured as a change in polarization state. From [21], modified.
. . . . . . . . . . . . . . . . . . . . . . . . .
47
3.8
Cross sectional TEM image and ellipsometry model results . . .
50
3.9
Annealing study - time scan - for a-Si:H/SiNx dual layer passivation performed at 300◦ C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.10 RSM experiment for PECVD SiNx optimization
. . . . . . . . .
51 53
3.11 Experiment runs - Points in the design space selected according to DOptimality criterion. Size of circles represents gas ratio R = 1 · · · 5 . . . . 4.1
54
Passivation with Hydrofluoric Acid - in-situ measurement of injection level dependent minority carrier lifetime immersed in 49% HF with QSSPC method in transient mode . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2
56
HF passivation stability - stability of HF chemical passivation upon exposure to atmosphere
. . . . . . . . . . . . . . . . . . . . . . . . . . .
57
4.3
Minority carrier lifetime - Response surfaces for R=1, R=5 and Po =20W 61
4.4
Minority carrier lifetime - Response surfaces for R=1, R=5 and Po =80W 61
4.5
Numerical optimization of response surface - deposition conditions associated with the peak point are Rg =8, Po =80W, T =250◦ C, P =500mTorr 63 xii
4.6
Refractive index nr at λ = 633nm - surface is described by two factors, Rg and T
4.7
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
65
Growth Rate [nm/min] - response surfaces for P=20W, P=50W, and P=80W . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
68
4.8
Growth Rate [nm/min] - response surfaces for R=1, R=3, and R=5 .
68
4.9
Annealing study - time scan - for a-Si:H/SiNx dual layer passivation performed at 300◦ C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
70
4.10 Annealing study - a-Si:H thickness dependence - for a-Si:H/SiNx dual layer passivation with thin (5nm) and thick (17nm) a-Si:H . . . . .
73
4.11 Annealing study - gas ratio dependence - for a-Si:H/SiNx dual layer passivation with SiNx gas ratios R = 1, R = 2, and R = 3 . . . . . . . . .
74
4.12 Annealing study - the effect of pre-deposition annealing - for aSi:H/SiNx dual layer passivation with SiNx gas ratios R = 3, T=180◦ C, Po = 20W , P = 1000mT orr . . . . . . . . . . . . . . . . . . . . . . . . .
75
4.13 SE measured values tan Ψ and cos ∆ - for a-Si:H/SiNx dual layer passivation annealed at increasing temperatures . . . . . . . . . . . . . . . .
77
4.14 MW-PCD surface scan of NSWHJ 452 sample - a-Si:H/SiNx (9nm/144nm) dual layer passivation with SiNx deposition conditions R = 8, T = 250◦ C, Po = 80W , P = 500mT orr . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1
79
BACH solar cell concept with dual layer a-Si:H/SiNx passivation - emitter and contacts are placed on the rear surface to avoid shadow losses 82
5.2
BACH solar cell concept with dual layer a-Si:H/SiNx passivation - emitter and contacts are placed on the rear surface to avoid shadow losses,
6.1
from [12] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
83
a-Si:H/SiNx Dual layer passivation
85
xiii
. . . . . . . . . . . . . . . . . .
A.1 Response Surface Methodology - experiment schematic diagram showing predictor factors (X), errors (), concomitant factors (Z) and observed responses (Y) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
88
B.1 Annealing study - boron doped a-Si:H structures - effect of annealing on p-a-Si:H/SiNx , p-a-Si:H/SiOx , p-a-Si:H/ONO . . . . . . . . . . . .
96
B.2 Annealing study - boron doped a-Si:H structures - effect of predeposition annealing on p-a-Si:H/SiNx , p-a-Si:H/SiOx ,p-a-Si:H/ONO . .
97
B.3 Annealing study - multilayer films - a-Si:H/SiNx , a-Si:H/SiOx , aSi:H/ONO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
xiv
98
1 Introduction
1.1
Motivation
Environmental sustainability calls for responsible use of resources available today without compromising the ability of future generations to meet their needs. Unfortunately, current resource consumption is not sustainable. The lack of sustainability is evident from our ever increasing dependence on limited fossil fuels. Highlighting our strong dependence on oil, was the peak oil price of $147.27 per barrel in 2008. Spiking oil price arising from the volatility of financial markets and diminishing supply of readily accessible reserves has sparked speculations of the next peak oil price reaching the $200 per barrel mark. Photovoltaics (PV) are one of the available technological options which shifts towards renewable and carbon dioxide free energy supply. Fundamentally, photovoltaic energy sources will yield a reduction energy costs with time instead of a guaranteed increase, as in the case of oil. This fundamental property alone makes photovoltaics an option worth pursuing. The potential of PV is immense. For decades, it was known that just a small percentage of the sun’s energy reaching the earth’s surface could supply the energy demands of 1
1.2 Background
Figure 1.1: Solar modules installations - historical development of World cumulative PV power installed in main geographies, adapted from [28] mankind several times over [17].
1.2
Background
According to the recently published data [28], global solar cell production in 2009 was estimated at 11.5 GW. This level of production output is equivalent to a 56% increase when compared to 2008 levels. Since the year 2000, the photovoltaic production increased at an annual rate ranging between 40-80%. The most rapid production growth over the last years has been attributed to China and Taiwan, which cumulatively account for about 50% of the global production levels today. The production demand has been sustained by a rapid increase in PV module installations world wide, as shown in the historic data in Figure 1.1. About 80% of the current PV production uses crystalline silicon (c-Si) technology. Manufacturers typically choose this approach due to the predictability of production line setup. Since the technology is relatively established, many turn-key solutions are available for production line start ups. 2
1.2 Background
Figure 1.2: Costs of a typical c-Si solar cell - crystalline silicon amounts to 75% of costs at the cell level and 50% of solar module, [18]
PV market goals and trends are centered around cost reduction. The costs of production of c-Si solar cells are strongly correlated to the material consumption per Watt. Since the majority of a typical c-Si solar cell cost is attributed to substrate material, Figure 1.2, the goal of cost reduction can be accomplished by both increasing the efficiency of a device, yielding higher power, and by reducing the mass of device substrate. Research efforts are driven towards decreasing the material consumption from 8g per Watt to as low as 6g per Watt [28]. The power produced by a solar cell is related to its efficiency, which is in turn governed by the recombination of photogenerated charge carriers in the bulk and at the surface where the latter dominates as the quality of the bulk material is improved (in sufficiently thin absorbers). The reduction of surface recombination losses is referred to as passivation. The significance of the surface recombination is due to two factors. Firstly, the reduction of material consumption is synonymous to wafer thickness reduction, which also increases surface-to-volume ratio. A larger relative surface area raises the importance of electronic passivation of surface recombination centres. Secondly, the diffusion lengths of the minority charge carriers in the high-efficiency Si solar cells are large and make the complete volume of the solar cell electronically active, making surface defect 3
1.2 Background sites available for recombination of photo-generated charge carriers. Surface passivation therefore plays a key role in improving solar cell efficiency at lower substrate thicknesses. To date, a significant amount of research has been carried out by the PV researches to address the need for better surface passivation for solar cell applications [3]. There are two fundamentally different means to address the issue of surface recombination losses: 1. Reduction of the density of surface states, 2. Reduction of the carrier concentration near the surface. Both of these strategies are typically incorporated in high-efficiency Si solar designs. The standard (and very effective) technique for the reduction of surface state density is the thermal oxidation of the silicon surface at high temperatures of over 1000◦ C. The second strategy is realized through a high temperature diffusion process used in the formation of an high-low isotype junction. The gradient in the dopant concentration gives rise to an electric field that can be configured to serve as a minority carrier mirror, that is, pushing the minority carriers away from the surface/interface. This reduction of carrier concentration at the surface effectively decreases the rate of recombination, given that the recombination rate is directly proportional to the defect density and minority carrier concentration. In the context of a high-efficiency Si solar cell based on a p-type absorber, a p+ p dopant profile is formed to obtain the so-called back surface field (BSF). However, variation in the dopant distribution is not the only way to form a desired electric field below the surface. The fixed, stable charges at the surface or in a top layer can promote field effect passivation. The stable charges can be formed in an insulator layer, such as thermally grown silicon dioxide SiO2 or silicon nitride SiNx deposited by means of the plasma chemical vapour deposition technique. A combination of thermal oxide passivation, a diffused p-n junction at the front surface, and a local BSF resulted in energy conversion efficiencies of 25% achieved by the 4
1.2 Background passivated emitter and rear locally diffused (PERL) solar cell design developed at the University of New South Wales [22]. The challenges in the design of solar cells and the approaches to the passivation are to achieve maximum performance with cost-effective processing. Presently, the designs drawing on the performance of the thermal oxide are constrained by the high temperature requirement for oxidation. The high temperature process expands the thermal budget of the cell production and forces the formation of metal contacts after the high temperature step so as to avoid shunting the p-n junction and degrading the carrier lifetime with potential diffusion of metal impurities. A significant improvement in the fabrication process was the introduction of SiNx as a passivating layer instead of SiO2 . The process temperature was cut back to below 400◦ C and the built-in positive charges in the SiNx layer enhanced the field-effect passivation of c-Si surface. Further, metal contacts are typically formed by silicon screening over the SiNx and subsequently firing rapidly thereby ensuring that the metal fingers appropriately diffuse through the SiNx and make contact with the underlying doped region. One solar cell design, drawing on the advantages of SiNx , was proposed at Institute f¨ ur Solarenergieforschung Hameln/Emmerthal (ISFH). Its conversion efficiency was shown to go up to 20.1% [26]. The other advantage of PECVD SiNx is the ability to adjust the refractive index in the range of 1.8 - 2.4, which makes it an effective antireflection coating because the tuning range can form a good impedance match between air and Si wafer and SiNx has a low extinction coefficient in the spectral range of most solar radiation. The thermal budget of cell processing is further decreased with hydrogenated amorphous silicon (a-Si:H) films deposited by PECVD at temperatures below 250◦ C. The a-Si:H films are very effective in reducing the surface state density, matching the performance of thermal oxidation. The excellent electronic passivation of c-Si by thin a-Si:H films has been explored in industry for the fabrication of high-efficiency a-Si:H/c-Si heterojunction solar cells. Sanyo’s HIT cell design is capable of 22% energy conversion 5
1.2 Background efficiency [59]. However, the drawback of using amorphous silicon as a passivation layer of the front surface of a solar cell is its significant absorption of the incident photons with energies above the bandgap. In a drive towards ultra-thin a-Si:H passivating layers, a good compromise between high transmission and good passivation was achieved for 6.5-10nm thick a-Si:H capped by SiNx by Plagwitz [44]. Some of the advantages of individual passivation methods can be combined when the passivating films are stacked together. The resulting multi layer structures have the flexibility for optimization for a specific application. However due to the associated interaction between layers, the process of optimization is not a trivial combination of properties of individual films. The dual layer structure consisting of a-Si:H and SiNx , which forms a cornerstone of this work, was first introduced by Bentzen et al. [7] who studied the effect of passivation and the distribution of hydrogen within the two layers. The improvement of passivation quality of the dual layer stack on p-type 2Ωcm wafer compared to single layer a-Si:H was attributed to the release of hydrogen toward the substrate, possibly passivating the dangling bonds in the sub-interface region. The reported surface recombination velocity (SRV) for 50nm thick a-Si:H layer was 7cm/s on 6Ωcm p-type Chochralski Si wafers. A big improvement in the reported passivation quality, reaching as low as 0.75±0.6 cm/s with the dual layer a-Si:H/SiNx stack featuring a 10nm thick a-Si:H layer on p-type 1.5Ωcm wafers was reported by Gatz et al. [49]. The dual layer structure was applied to n-type wafers by Tucci and Serenelli [60]. Their dual layer passivation has shown an initial lifetime of 720µs and the SRV of 36 cm/s on n-type 1Ωcm, which was shown to improve to around 1300µs during annealing. The characterization of these samples by infrared spectroscopy did not reveal any changes in recorded spectra, implying annealing up to 500 ◦ C did not cause structural changes within the films. A recent publication by Focsa et al. reported SRV of 10.7 cm/s on n-type 1 − 10Ωcm FZ c-Si wafers. The report has shown that the deposition of the dual layer structure in two steps, separated by an 6
1.3 Research Questions exposure to air enhances the passivation quality and suggested that presence of native oxide on a-Si:H layer may contribute to that improvement. A graphical summary of the results reported in the literature is shown in Figure 2.5 on page 27. While there are many dual layer passivation studies available on p-type c-Si, there are only a few studies focusing on the n-type substrates. In addition, the flexibility of plasma deposition processes has not been systematically explored or optimized in the context of a-Si:H/SiNx dual layer passivation. Together, these two questions create an opportunity for an effective research study, the results of which can be applied to high-efficiency advanced heterojunction solar cell concepts, such as the back amorphouscrystalline silicon heterojuction (BACH) solar cell, discussed in Chapter 5.
1.3
Research Questions
This work explores a method for passivation of n-type crystalline silicon surface using a dual layer amorphous silicon structure consisting of a thin hydrogenated amorphous silicon (a-Si:H) layer and amorphous silicon nitride (SiNx ) layer. A sought advantage of this approach is maintaining and enhancing the high quality passivation properties of thin a-Si:H and improving the anti-reflective properties of solar cell devices with SiNx film. While many reports discuss optimal conditions for SiNx deposition as a single passivating layer deposited using radio frequency plasma enhanced chemical vapour deposition (RF PECVD), no systematic study has been reported on the application of SiNx as a capping layer for a-Si:H deposited using the DC saddle field (DCSF) PECVD system. In order to gain insight into significant parameters of radio frequency plasma enhanced chemical vapour deposition system influencing the passivating quality of multilayer passivating films, an optimization study was conceptualized using a statistical response surface methodology (RSM) and a design of experiments (DOE) approach. 7
1.4 Thesis Outline The deposition conditions from RSM, which resulted in best performing passivation multilayer films, were singled out for a study of the evolution of multilayer film properties upon annealing. Characterization tools examining the evolution of carrier lifetime and optical properties were used to gain insight into the passivation mechanism of dual layer structures. The best performing passivating films resulting from the optimization study are proposed for Back Amorphous-Crystalline silicon Heterojunction (BACH) solar cell [25], [34].
1.4
Thesis Outline
The research work reported herein is divided into the following chapters: Chapter 2 reviews the physics of carrier recombination at the surface and within the bulk of crystalline silicon. These concepts are applied in the discussion of the mechanisms of surface passivaton. Building on the theoretical approaches, the chapter will present the state of art results in the field of surface passivation. Chapter 3 focuses on the experimental details, which begin with a discussion of synthesis techniques of amorphous silicon films. The discussion covers sub topics of pre-deposition treatment methods, methods of deposition of passivating films and postdeposition steps. Later in the chapter, characterization methods used to assess the quality of passivating films are described. Chapter 4 presents the results of experimental work. The opening section of the chapter presents the results of in-situ assessment of chemical passivation of bare c-Si. The chapter then focuses on the results from the RSM study and shows the surface plots of measured responses. The RSM plots yield deposition conditions for best performing passivation layers through numerical optimization techniques. The predicted values of responses at these deposition conditions are compared against experimental results. 8
1.4 Thesis Outline The chapter moves to discuss the effect of post-deposition annealing on the passivation quality and the structure of deposited films. Later in the chapter effects of the a-Si:H thickness and the precursor gas ratio on annealing behaviour of the dual layer a-Si:H/SiNx passivating structure are shown. Chapter 5 presents the details of BACH solar cell concept and shows the potential conversion efficiency of the BACH cell with the use of optimized a-Si:H/SiNx dual layer passivation developed in this work. A summary of the work and possible future directions are presented in the concluding chapter.
9
2 Mechanisms of Surface Passivation
2.1
Definitions
Carrier recombination in semiconductors is most commonly defined as a transition of an electron from the conduction band to valence band, a process which is also referred to as electron-hole annihilation. Both electrons and holes in the conduction and valence bands, respectively, are free charge carriers. Their concentrations are dependent on incident illumination, temperature and dopant impurity concentration in a semiconductor material. For a non-illuminated semiconductor in thermal equilibrium, the concentration of electrons in the conduction band is n = n0 and holes in the valence band p = p0 . These concentrations remain constant due to the equilibrium between generation and recombination processes. These concentrations for an intrinsic semiconductor (where carriers are present only due to thermal excitations) can be calculated from: Z
∞
n0 =
Z
∞
dn0 = Ec
f0 (E)gc (E)dE
(2.1)
Ec
here, in Equation 2.1, f0 (E) represents the equilibrium Fermi distribution function and gc (E) is the density of states. In the case of electrons, the integration is from Ec the conduction band edge to the top of the conduction band (denoted here as infinity owing 10
2.1 Definitions to exponential character of the integrand). The integral in 2.1 is evaluated to yield Ec − Ef (2.2) n0 = Uc exp − kT where Uc is the equivalent density of states at band edge, Ef is the Fermi energy, k is the Boltzmann constant, and T the temperature. For a doped semiconductor, the band gap Eg and position of Fermi energy Ef are dependent on the dopant impurity concentration. However, in all semiconductors the product n0 p0 is independent of the Fermi level or impurity content and is only a function of the effective masses and the band gap. For a given semiconductor material both of the latter variables are fixed; hence, the product n0 p0 is only a function of temperature. This product is referred to as mass-action law that governs the relative concentration of electrons and holes in a semiconductor. In the case of intrinsic semiconductor, the relative concentrations must be equal. n0 = p0 = ni (T )
(2.3)
In the equation 2.3 ni is the effective intrinsic carrier concentration; in silicon at T=300 K the commonly accepted value1 of ni is approximately 1.5 × 1010 cm−3 According to these relations, the concentration of carriers in an intrinsic semiconductor is a function of temperature and band gap, increasing with temperature and decreasing with an increase in Eg . For n-type silicon with a dopant concentration ND = 1 × 1016 cm−3 , used for carrier lifetime analysis in this work, n0 = 1 × 1016 cm−3 and p0 = 2.25 × 104 cm−3 . In general, if one assumes full ionization of impurity atoms,
1
n2i for n-type semiconductor ND n2i for p-type semiconductor = NA
n0 = ND , p0 =
(2.4)
p0 = NA , n0
(2.5)
Various references list slightly different values of intrinsic carrier concentration in silicon at room
temperature. In general, they are between 1 × 1010 and 1.5 × 1010 cm−3
11
2.1 Definitions Under illumination, the incident photons of energy equal to or exceeding the band gap energy of the semiconductor, are absorbed and as a result, an excess concentration ∆n of electron-hole pairs is created. The excess electron-hole pairs thus created contribute to the conductivity of the semiconductor crystal, which is increased under illumination. This phenomenon forms a basis for some of the characterization techniques used in this work which are described in Section 3.2.1. The excess carrier concentration decays after the illumination has been turned off with a net recombination rate, U: U (∆n(t), n0 , p0 ) = −
∂∆n(t) ∂t
(2.6)
Since different recombination mechanisms contribute to the net rate, the expression 2.6 is rewritten as follows: U (∆n(t), n0 , p0 ) =
X
Ui (∆n(t), n0 )
(2.7)
i
If we treat carrier recombination rate to be independent of excess carrier concentration, we can solve equation 2.6 as ∆n(t) = ∆n(0)exp(− τt ). This time constant τ is referred to as carrier lifetime. Using Equation 2.6 and the preceding exponential decay solution, we can express the effective carrier lifetime as ∆n(t) = τ= U (∆n(t), n0 , p0 )
X −1 1 . τi i
(2.8)
Equation 2.8 shows that the time constant τ for the net recombination process is a sum of time constants τi from individual recombination rates A mobile charge carrier will travel a certain distance in the time τ before it recombines. This distance is referred to as the diffusion length and it is expressed as √ L=
Dτ
(2.9)
D is an impurity concentration dependent diffusion constant. Both relations for carrier lifetime 2.8 and diffusion length 2.9 are functions of ∆n. Therefore, when quoting lifetime or diffusion length parameters it is important to state the excess carrier injection level. 12
2.2 Bulk Recombination
2.2
Bulk Recombination
In a semiconductor the term bulk refers to a region in the base material which is free of valence and conduction band bending, typically associated with a disturbance in the periodicity of crystalline structure, such a disturbance happens near a surface. Carrier recombination in bulk occurs via two intrinsic and one extrinsic mechanism. Intrinsic mechanisms include radiative band-to-band and non-radiative Auger recombination. The extrinsic recombination mechanism is associated with the presence of impurity dopant atoms that create an additional path for recombination via energy levels within the band gap. This mechanism is described by Shockley-Read-Hall (SRH) theory. The effective carrier lifetime in bulk material τb is described as an accumulation of individual lifetimes associated with each of the three recombination mechanisms according to equation 2.10 1 1 1 1 = + + . τb τrad τAuger τSRH
2.2.1
(2.10)
Radiative Recombination
Radiative recombination occurs when an electron ‘falls’ from the conduction band to the valence band and recombines with a hole, thus annihilating the pair is accompanied by the release of a photon that carries away the excess energy. This process is the reverse of absorption. The probability of this process is lower in an indirect semiconductor when compared to a direct semiconductor material because there is a need for an additional phonon to satisfy the momentum conservation in the crystal. Therefore one could expect larger lifetime values for indirect semiconductors however this has not been observed. Additionally, this process does not explicitly take into account the observed dependency on doping concentration. Both of these facts suggest that radiative recombination plays a secondary role in establishing the lifetime of charge carriers. The total radiative re13
2.2 Bulk Recombination combination rate is proportional to the product of concentrations of holes and electrons and the net recombination rate is given by the following expression (excluding the non illuminated case): Urad = B(np − n2i ),
(2.11)
where B is the radiative recombination coefficient. The temperature dependence of this coefficient has been evaluated at room temperature for silicon B = 9.5 × 10−15 cm3 s−1 [50]. Combining Equations 2.11 and 2.8, the effective lifetime τrad for radiative recombination process is τrad =
2.2.2
1 B(n0 + p0 + ∆n)
(2.12)
Auger Recombination
Another method of band to band recombination in bulk semiconductor materials is described by the Auger effect. The distinguishing fact is that the excess energy from recombination of electron and hole is not carried away by a photon, as is the case with radiative recombination, but instead it is transferred to another carrier. This second carrier, which is either an electron in the conduction band (eeh - process) or a hole in the valence band (ehh - process), then dissipates the extra energy through lattice collisions and returns to lower energy level at the band edge. The rates for two Auger processes should depend on concentration of reaction species; thus, for example the electron-electron-hole (eeh) process will be proportional to n2 p. The overall net reaction, having subtracted the non-illuminated case, becomes UAuger = Cn (n2 p − n2i n0 ) + Cp (np2 − n2i p0 )
(2.13)
The coefficients for eeh and ehh processes, Cn , Cp , have been determined for higher 14
2.2 Bulk Recombination doping concentrations of > 1018 cm−3 [15]. However, these coefficients appear to be off in the dopant range of 1016 − 1018 cm−3 by an order of magnitude [36]. The difference in rates established by Equation 2.13 and experiment has been attributed to Coulomb factors that enhance the Auger recombination lifetime [23]. At higher doping levels, the Coulomb factors do not play as much of a role due to a large amount of scattering within the lattice. Additionally, phonon-assisted processes have been used to explain the variation of recombination rates with doping concentration. Due to various contributing factors to Auger lifetime, a complete quantitative study is difficult. Instead, Cuevas and Kerr have devised an empirical model that fits the observed experimental behaviour in the doping concentration range between 1013 − 1020 cm−3 [33] UAuger = np(1.8 × 10−24 n0.65 + 6.0 × 10−25 p0.65 + 3.0 × 10−27 ∆n0.8 ). 0 0
(2.14)
The intrinsic carrier lifetime in silicon given in Equation 2.8, can now be re-written by substituting in Equations 2.12 and 2.14
τint =
∆n 1 + 0.65 −24 B(n0 + p0 + ∆n) np(1.8 × 10 n0 + 6.0 × 10−25 p00.65 + 3.0 × 10−27 ∆n0.8 ) (2.15)
Figure 2.1, shown below, is a plot of Equation 2.1 for a range of excess carrier densities for n-type silicon with two different doping concentrations. Equation 2.1 represents the upper limit for the carrier lifetime in crystalline silicon.
2.2.3
Recombination via Defects
A different type of recombination mechanism is recombination via defects. It is an extrinsic effect that arises due to lattice defects, such as dislocation or impurity (dopant atoms). The lattice defects create additional energy levels typically within the forbidden 15
2.2 Bulk Recombination
Figure 2.1: Intrinsic carrier lifetime in n-type silicon - ND = 1.53 × 1015 (3 Ωcm) and ND = 1.45 × 1014 (30 Ωcm) band. The presence of these energy levels enhances the carrier recombination. The model for this recombination mechanism was developed by Shockey and Read [52] and Hall [24]. Their proposition was to determine the recombination rate USRH due to Nt defects at an energy level Et , material and excitation parameters on the basis of statistical considerations of the four fundamental processes, shown in Figure 2.2, and enumerated below: 1. an electron is captured by a free/unoccupied defect energy level 2. an electron is emitted from an occupied defect energy level into the conduction band 3. a hole is captured by an occupied defect energy level 4. a hole is emitted from an occupied defect energy level into a free state in the valence band 16
2.2 Bulk Recombination
Figure 2.2: Electron energy band diagram showing four interaction mechanisms of free carriers with a defect level: electron emission(1), electron capture (2), hole capture (3), hole emission (4) The net Shockley-Read-Hall (SRH) recombination rate of excess charge carriers: USRH =
(np − n2i ) τn0 (p + p1 ) + τp0 (n + n1 )
(2.16)
where the capture time constants are defined as 1 for electrons Nt σn vth,n 1 for holes = Nt σp vth,p
τn0 =
(2.17)
τp0
(2.18)
and the densities of occupied trap states are
n1 p1
Et − Ei = ni exp for electrons kT Ei − Et = ni exp for holes kT
where,
Ei is intrinsic Fermi level vth is thermal velocity, vth,p ≈ 1.7 × 107 cm/s and vth,n ≈ 2.0 × 107 at 300K [44] 17
(2.19) (2.20)
2.2 Bulk Recombination
Figure 2.3: Bulk recombination mechanisms - Excess carrier density dependence of recombination mechanisms in the bulk of n-type silison ND = 1.53 × 1015 (3 Ωcm). SRH assumes defect levels of Fe in Si, with Nt = 1 × 1010 cm−3 σn , σp is the capture cross section for electrons and holes, respectively
τSRH =
τn0 (p + p1 + ∆n) + τp0 (n + n1 + ∆n) n0 + p0 + ∆n
(2.21)
SRH formalism Equation 2.21 completes the discussion of bulk recombination mechanisms. Figure 2.3 summarizes the individual effects of aforementioned recombination mechanisms as a function of the excess carrier density as well as the cumulative effect on the bulk carrier lifetime, following the Equation 2.10. Since the bulk lifetime assumes no surface effects, its plot establishes the best case scenario for surface passivation. Therefore, high bulk lifetime material can be used to benchmark the performance of various passivating schemes. 18
2.3 Surface Recombination
2.3
Surface Recombination
Physical boundaries of a crystal terminate its translational and rotational symmetries. When a boundary divides a crystalline semiconductor from a gas or vacuum it is referred to as a surface. If instead of a gas or vacuum the there is another solid or a liquid, the boundary is referred to as an interface. The symmetries in a periodic crystal lattice allow for some simplifications in the analysis of a crystal structure, however in the absence of symmetries near a surface and the presence of amorphous passivating layers the analysis becomes more complex. As a result a variety of factors affect the recombination at the surface which include recombination through surface states, enhanced recombination due to surface band bending associated with appropriately doped or passivating films, all of which collectively are referred to as effective surface recombination.
2.3.1
Extended SRH formalism and Surface Recombination Velocity
In periodic infinite crystals the allowed energy states for an electron lie in bands which are separated by forbidden gaps. However, imposing a boundary condition of a free surface in a one dimensional model of a semi-infinite crystal results in the creation of surface states. Surface states are energy levels which are situated between the valence and conduction bands of a periodic semiconductor. These surface states are active recombination centres and the phenomenon is referred to as surface recombination. The surface states are continuously distributed within the band gap [57], [51] and the appropriate approach to describe surface recombination is to extend the SRH formalism, described by Equation 2.16, by integrating over all energies within the band gap: 19
2.3 Surface Recombination
Us = (ns ps −
n2i )
Z
Ec
Ev
σn−1 (Et )(ps
vth Dit (Et )dEt + p1 ) + σp−1 (Et )(ns + n1 )
(2.22)
This equation is similar to Equation 2.16, with additional parameters Dit being the interface trap density, ns and ps being electron and hole concentrations at surface and σn (Et ) and σp (Et ) are energy-dependent capture cross-sections. The recombination rate at the surface is not measured in terms of a time constant, as was done with bulk material. Given that, Dit has units of cm−2 , this implies that Us has units of cm−2 s−1 and the quantity σvth Dit has units of velocity instead of reciprocal time. Therefore, similarly to Equation 2.8, surface recombination velocity (SRV), S, is defined as S=
Us ∆ns
(2.23)
SRV is a useful parameter to evaluate the quality of the passivation mechanism. Higher SRV implies quicker recombination of carriers at the surface. Therefore, the goal of high quality surface passivation is to minimize SRV. However, due to the large number of energy-dependent variables in Equation 2.22, it is possible for SRV to change with the injection level. The injection dependence of SRV has been discussed in detail by Aberle [2]. In the more general case, a surface space-charge region will be present close to the surface. This is typical when there is a passivating layer present on the surface. Since two different substances come into contact, there will be some band bending in the vicinity of the interface due to to the difference in Fermi levels of the two substances and due to the presence of charge in the passivating layers. For example, PECVD deposited a-SiNx :H films have a positive fixed charge density, as was shown by [38], [1]. In order to account for the band bending, it is proposed to use define SRV Seff by combining together the recombination effects from the surface states and space charge region. The Seff can be thought of as a shift of the semiconductor boundary into the 20
2.3 Surface Recombination bulk region [19] to the edge of the surface space charge region. Additionally, the presence of two dissimilar materials offers band offset, which describe how the valence and conduction bands align when the two materials come into contact. The conduction band offset of 2.2eV is nearly twice as much as valence band offset of 1.2eV [55]. The band offsets confine more carriers to a-Si:H layer and the distribution of localized interface states and the band bending are the main factors that influence the behaviour of the interface. Thus, it was shown that there exists an electron accumulation layer in a-Si:H at the interface with Si3 N4 [55].
Seff =
2.3.2
Us ∆n(xd =d)
(2.24)
Effective carrier lifetime
The excess carrier lifetime τeff is defined in terms of the bulk lifetime τb and effective surface recombination velocity Seff [2]; the expressions presented below are for n-type material: 1 1 = + α12 Dp τ τ eff b α1 W Seff tan = 2 α1 D p
(2.25) (2.26)
where, α1 is the solution to the transcendental equation, Dp is the diffusion constant for holes and W is wafer thickness. For sufficiently small Seff and symmetric structure, where both surfaces having the same SRV, [54] applied the following simplification: 1 1 2Seff SW 1 = + , which holds for < τeff τb W Dp 4
(2.27)
Equation 2.27 is accurate for Seff < 300cm/s, Dp = 30cm2 /s and W = 250µm to within 4%, as was shown in [54]. This relationship will be used in the following sections. 21
2.3 Surface Recombination
2.3.3
Surface Passivation Mechanisms
From extended SRH formalism to encompass surface states, Equation 2.22, it is evident that there are two types of variables that influence the rate of carrier recombination at the surface or interface:
• Interface trap density Dit Rate of recombination is proportional to the density of defects at the surface. In principle, it is necessary to saturate the dangling bonds that exist at the surface, thus reducing the number of defects. This principle is used in passivation with dielectric films like SiO2 and SiNx and also with hydrogenated amorphous silicon. Another example is chemical passivation with hydroflouric acid (HF). In all of these examples, hydrogen or oxygen atoms from the passivating film or medium are used to saturate the dangling bonds.
• Carrier densities at interface (ns and np ) Equation 2.22 shows a direct dependence of the rate of SRH recombination on the concentration of carriers at the interface or surface. The much desired reduction of carriers at the interface can be accomplished by fixed built-in charges in SiO2 or SiNx films. This effect is called field effect passivation. Another approach is to form the so called back surface field BSF or front surface field effect FSF by introducing a sufficiently highly doped surface layer which serves as a minority carrier mirror.
Technologically, the passivating layers often employ both principles simultaneously. For example, silicon nitride film forms effecting an interface passivation layer which includes field-effect passivation due to presence of the built in positive charges. 22
2.4 State of the Art Passivation Mechanisms
2.3.4
BACH solar cell efficiency dependence on front SRV
Back Amorphous-Crystalline silicon Heterojunction (BACH) solar cell concept, proposed by Hertanto, Liu et al. [25] will be discussed in more detail in Chapter 5. The assessment of influence of front SRV on BACH solar cell performance has been made in a simulation study by Chowdhury, et al [12]. The study was carried out with the Sentaurus simulation software package. The simulation results presented in Figure 5.2 show a strong dependence of cell efficiency η on both wafer thickness W and Seff . The results show a dominant role of the surface recombination with thinner wafers.
2.4
State of the Art Passivation Mechanisms
2.4.1
Thermal SiO2
Thermally grown silicon dioxide, SiO2 , is the state of art method for surface passivation of c-Si. The synthesis process sequence is independent of doping type and level, which is a very convenient feature that guarantees high quality passivation every time. The best passivation results, to the best of our knowledge, are presented by Kerr and Cuevas [33]. They have reported τeff = 29ms on 90Ωcm Float-zone (FZ) n-type h100i c-Si and τeff = 32ms on 150Ωcm FZ p-type h100i c-Si. The drawback of this high temperature fabricated passivation material is evident in the fabrication of solar cell devices:
1. High reflection losses due to small refractive index of 1.46. 2. High temperature synthesis process (in the range of 1000◦ C.) poses a number of problems: 23
2.4 State of the Art Passivation Mechanisms • The thermal budget associated with the process are high, which adversely affect the $/Watt ratio.
• High temperatures increase the rate of diffusion of impurities. In particular metals as well as surface defects, due to their high diffusion coefficient at elevated temperatures , will migrate into the bulk. Diffusion of impurities can be detrimental to the bulk lifetime. As a result, high purity electronic grade process gases as well as high-purity high-temperature chambers are required for production, which further contribute to the total costs.
2.4.2
PECVD SiNx
Amorphous hydrogenated silicon nitride SiNx layers are deposited by PECVD at temperature of around 400◦ C. In addition to a lower thermal budget for production, SiNx layers are used as antireflection coatings due to refractive index in the range of 2.0-2.4. However, SiNx is shown to have large built in positive charges which consequently form an inversion layer at the surface of p-type Si. The solar device efficiency is reduced due to shunting of inversion layer and base contacts - which effectively enhances recombination of carriers. However, in n-type silicon, positive charges in the film create field effect passivation to shield minority carriers away from the recombination active interface area. To our knowledge, the best passivation results with SiNx were reported by Lauinger et al., [37] with effective surface recombination velocity Seff of 4cm/s on low-resistivity (1.5 Ωcm) FZ p-type c-Si using microwave-excited remote PECVD technique. 24
2.4 State of the Art Passivation Mechanisms
2.4.3
Hydrogenated amorphous silicon a-Si:H
Surface passivation by intrinsic amorphous silicon a-Si:H deposited by DCSF PECVD at temperatures below 250◦ C combines the advantages of good interface defect passivation of SiO2 and low temperature process. Effective surface recombination velocities of around 2-5cm/s were reported in literature [14]. However, a-Si:H films absorb incident light at wavelengths below 730nm or above the a-Si:H bandgap Eg = 1.7eV [58]. Therefore, solar cell device performance (due to short circuit current density losses JSC )is hindered by increasing thickness layers of a-Si:H. Excellent passivation using thick (30 nm) i-a-Si:H layers has be reported giving an effective lifetime of more than 4 ms by Bahardoust et al. [5].
2.4.4
Dual layer passivation
A good balance among the advantages of three passivating films described above is achieved in a combination two or more films into one passivating structure. The main focus of this work is the study of a-Si:H/SiNx dual layer passivation, illustrated in Figure 2.4. The advantages of this approach are:
1. a-Si:H forms high quality interface passivation 2. SiNx provides field effect passivation due to the built in positive charges and serves as an antireflection coating with index of refraction n ≈ 2.0 3. a-Si:H/SiNx dual layer stack allows for thinner a-Si:H layer, which minimizes the absorption of the incident light. In dual layer configuration most effective a-Si:H thicknesses are 6.5-10nm [44]
25
2.4 State of the Art Passivation Mechanisms
Figure 2.4: a-Si:H/SiNx dual layer passivation - combines excellent interface passivation of a-Si:H with field-effect passivation and anti reflection properites of SiNx The reported SRVs are graphically summarized in Figure 2.5. The reports predominantly concentrate on p-type c-Si. The reported film synthesis techniques are typically rf-PECVD or microwave-excited remote PECVD deposition techniques. To date, no systematic optimization of SiNx capping layer for dual layer passivation of n-type c-Si has been reported. This study sets out to develop the optimal rf PECVD SiNx film deposition conditions for the use as a capping layer on DCSF PECVD intrinsic a-Si:H.
26
2.4 State of the Art Passivation Mechanisms
Figure 2.5: a-Si:H/SiNx Dual layer passivation - summary of SRV values published in the literature
27
3 Surface passivation by multilayer amorphous films
3.1
Synthesis techniques for amorphous films
Plasma enhanced chemical vapour deposition (PECVD) is an established technique for deposition of insulating films such as silicon nitride and silicon oxide. It is also widely used for deposition of hydrogenated amorphous silicon, by the dissociation of silane gas SiH4 . PECVD technique is favoured for lower process temperatures. PECVD is able to sustain film deposition by adding electrical energy into the CVD environment, which effectively substitutes some of the thermal energy. The plasma discharges are initiated and sustained by electric fields which are produced by either direct current (dc) or alternating current (ac) power supplies. The ac frequencies of operation can range from 100kHz to 13.56MHz in the radio frequency (rf) range. The two important features of a plasma reactor are to generate reactive neutral species and ions; and second, to make provisions for effective transport of these species 28
3.1 Synthesis techniques for amorphous films to the surface for film growth. A specified power supply provides the excitation energy to the glow discharge (named this way because it typically emits some light), where the precursor gases dissociate and generate reactive species. In this planar reactor configuration there is a a region of space free from a glow discharge, which is characterized by a strong electric field that accelerates the ions towards the substrate. The plasma deposition reactors are built to vary temperature, chamber pressure, precursor gas flow ratio and total gas flow volume and plasma power. Power supply frequency also has an effect, as was summarized by Lauinger [37], but the variation of this parameter is limited by FCC1 regulation. The aforementioned variables alter the rates of dissociation, gas phase and deposition reactions, which lead to significant variation of the properties of deposited materials. Due to the complexity of plasma processes, formation of accurate models of deposition is difficult. In the absence of accurate model predictions, the options for a systematic study of PECVD films are limited to experimental analysis and empirical approximations. In general, supplied power is related to the rate of dissociation of precursor gases; chamber pressure influences the frequency of collisions between gas molecules; further enhancing gas dissociation is the gas flow rate, which establishes the mean time that the gas spends in the reaction volume, and temperature significantly affects the surface reaction kinetics, necessary for film growth.
3.1.1
Cleaning and pre-deposition chemical treatments
Preparation of the surface prior to deposition has a great effect on the surface passivation properties. Surface preparation is embodied in removal of surface contaminants and surface roughness. The following contaminants affect the surface 1
Federal Communications Commission
29
3.1 Synthesis techniques for amorphous films properties:
1. Particles - disturb PECVD film growth 2. Metals - degrade minority carrier lifetime 3. Organics - serve as nucleation sites for non uniform PECVD film growth, diminish adhesion of PECVD films 4. Native oxide - gives rise to poor quality film, degrades interface properties 5. Micro-roughness - reduces surface mobility of growth species, which negatively affects the uniformity and consistency of PECVD films
Transition metal impurities are particularly detrimental because they significantly affect the properties of silicon by acting as charge carrier recombination centres. Even trace amount of impurities on the order of parts-per-trillion within active region can affect cell performance [39]. In solar cells the active region is the entire wafer including the surfaces; therefore wafer cleaning is an essential prerequisite for good device performance. A thermodynamic model for surface defect formation has been proposed by Reddy et al [45]. Figure 3.1 shows a comparison of electron energies of Si wafer (−4.52eV ) and electron energies of terminating groups. The difference between them constitutes the driving force for defect formation. Thus, for redox potential below Ef Si = −4.52eV an electron will be transferred to the metal ion, which will form a bond at the surface. This is why Cu is considered among the most critical metal impurities for Si devices. One of the most successful and popular wet chemical surface conditioning methods at the RCA Standard clean. It was developed in Radio Corporation of America in 1970 [32]. RCA clean consists of two steps, standard clean 1 (SC-1) and standard clean 2 30
3.1 Synthesis techniques for amorphous films
Figure 3.1: Thermodynamic Model for surface defect formation - from Reddy et al. [45] (SC-2). In SC-1 ammonium hydroxide, hydrogen peroxide and water are mixed in proportions NH4 OH : H2 O2 : H2 O = 1:1:5. This step is designed to solvate and break down by oxidation of organic-type molecules and particles on the surface. The ammonia NH3 serves as a mild oxide etchant, the hydrogen peroxide serves as a powerful silicon oxidizer. The peroxide continuously grows oxide only in areas in which the silicon is bare which results in a continual availability of oxide for the ammonia to remove. Some more recent cleaning approaches[30] added megasonic vibration energy to SC-1, which facilitate mechanical removal of organics. The megasonic acoustic cavitation is excited parallel to the wafer surface direction by means of acoustic waves, which are in turn produced by vibrations of piezoelectric crystal at frequencies between 500 and 2000kHz [6]. This improvement of the SC-1 recipe targets the removal of particles of 0.1 to 10µm in size by providing an external force to the particles to move them away from the wafer surface. A higher concentration of ammonium hydroxide is associated with the increase of surface roughening, which later affects the growth of thin passivating films. The oxide layer, grown in SC-1, is then etched in HF in the second step of RCA clean. In the third step, named SC-2, hydrochloric acid, hydrogen 31
3.1 Synthesis techniques for amorphous films peroxide and water are mixed in relative proportions of HCl : H2 O2 : H2 O = 1:1:6 by volume. The wafers are immersed in the solution which is at 70◦ C for 5 - 10 minutes. This step is designed to remove metal impurities not desorbed by SC-1 , alkali ions, hydroxides insoluble by NH4 OH, such as Al(OH3 ) from the surface by turning them into volatile metal chloride complexes, which do not adhere back to the surface [31].
3.1.2
a-Si:H and deposition technique using DCSF
a-Si:H properties
Unlike the perfectly ordered system of atoms in c-Si, amorphous silicon (a-Si) has a structure of a continuous random network of four-fold coordinated silicon atoms. This random structure features a deviation of the bond angle from the perfect tetrahedral arrangement of c-Si and varying bong lengths. In addition, the random network structure is associated with the missing bonds (dangling bonds), terminated with none, one or two electrons, corresponding to defect states D+ , D0 and D- . The random network does indeed have a band gap, as was shown by Weaire and Thorpe [62]. The density of dangling bonds in pure a-Si is on the order of 1020 cm−3 . However, hydrogenation of a-Si by about 10 atomic % reduces the number of defects to 1015 cm−3 [29]. Resulting material, hydrogenated amorphous silicon (a-Si:H) can be used in opto-electronic applications. Hydrogen plays a vital purpose of passivating the dangling bonds and enhancing optical and electronic properties of the material. Figure 3.2 shows the density of states in hydrogenated amorphous silicon [55], which features elements common to various amorphous semiconductors. The diagram labels the gap states, band tail states and extended states. The gap states, or deep defects, are associated with defect states D+ , D0 and D- . The band tails are observed due to strained Si-Si bonds and extended states arise from the Si network. The presence of hydrogen in a-Si 32
3.1 Synthesis techniques for amorphous films
Figure 3.2: Density of states in a-Si:H- adapted from [55] allows for creation of p- and n-type material by means of substitutional doping by Boron and Phosphorous, respectively. The process of doping is easily controlled through the ratio of silane SiH4 and diborane B2 H6 or phosphine PH3 . This flexibility is fully utilized in the fabrication of BACH solar cell.
DCSF PECVD System
The configuration of DCSF PECVD chamber used for a-Si:H deposition in this work consists of four parallel electrodes, Figure 3.3. Two of the central electrodes are made semitransparent (by means of using a wire mesh). The electrodes are held at ground potential with an exception to anode, which is powered by a dc power supply to dissociate the precursor gas mixture. The substrate is placed on upper electrode, which features a heater. The chamber is fitted with calibrated thermocouples for substrate temperature monitoring and a high vacuum gauge. This geometry in comparison to a dc diode significantly increases the probability of impact and other forms of precursor 33
3.1 Synthesis techniques for amorphous films
Figure 3.3: DCSF PECVD system - used for depositing a-Si:H films
gas dissociation. This increase is attributed to the higher density and larger mean free path of the electrons oscillating in the field. Since cathode is semitransparent, the activated gas species can travel towards the substrate surface, which is situated outside of the plasma region. Ions and radicals move towards the substrate by the electric field and diffusion, respectively. Silane, hydrogen, diborane, phosphine, nitrogen and inert gases are available through a common gas manifold connected to the chamber. The gases are metered into mixing bottles to a desired pressure and composition and subsequently metered out through a mass-flow controller during a given deposition. The pressure in the chamber is set through a pressure transducer feedback control of a motorized butterfly valve between the chamber and the vacuum system. A turbomolecular pump backed by an oil pump provides evacuation of the deposition facility. A SiH4 decomposition system, positioned between the turbomolecular and mechanical vacuum pumps, is used to decompose unspent gases from the deposition chamber prior to venting to the stack. 34
3.1 Synthesis techniques for amorphous films In this tetrode DCSF system configuration the film growth environment from the plasma region proper and thus permitting independent control of ion energy, ion density and flux at the substrate. The procedure for high quality a-Si:H film deposition is presented below:
1. Wafer preparation - the depositions were carried out on full 100 mm diameter wafers. The sealed wafer box was opened in an N2 glove box environment. The wafer was dipped into 5% diluted HF for 60 seconds prior to the deposition on the first side and subsequent 30 second dip prior to the deposition on the second side. No additional cleaning was required because the wafers were received pre-cleaned from the manufacturer. Independent tests confirmed the quality of the passivating film was not improved with additional of RCA cleaning of the wafers.
2. Chamber conditioning - the chamber was coated with intrinsic a-Si:H under the standard deposition conditions outlined in Table 3.1. Chamber coating step has shown to improve the quality of the film. The duration for this deposition was 12 min.
3. Deposition step - before commencing the deposition, the sample was pre-heated for 30 min in a nitrogen environment (to facilitate convective heat exchange). The deposition was then commenced for 10 min. The resulting a-Si:H film thickness remained below 10nm.
The conditions and steps in a-Si:H deposition were kept constant in this study at the settings summarized in Table 3.1, which have been previously determined, [5]. The a-Si:H films deposited with these conditions proved to be effective for the passivation of c-Si. 35
3.1 Synthesis techniques for amorphous films
Parameter SiH4 Gas flow
Setting 30 sccm
Substrate temperature
170 ◦ C
Heater temperature
320 ◦ C
Preheat time Operating pressure Anode voltage Anode current
30 min 160 mTorr 600-700 V 34.5 mA
Table 3.1: DCSF Parameter settings - a-Si:H deposition parameters remained constant in this study
3.1.3
SiNx deposition using rf PECVD
Fundamental Properties of PECVD silicon nitride films
Chemical, optical and electronic properties of silicon nitride films strongly depends on the method of deposition. The most common method of preparation is deposition of amorphous silicon nitride films from gas phase reactants by means of PECVD. Typically, this method uses hydrogen-containing reactants (in our case ammonia gas, NH3 ) in addition to silane diluted in nitrogen. The resulting hydrogenated amorphous silicon nitride, a-SiNx :H (it is more commonly referred to in literature SiNx ) is non-stoichiometric, with varying hydrogen, nitrogen and silicon content. The non-stoichiometry of the film is denoted with ‘x’ as the value of the ratio of nitrogen to silicon. Experimentally confirmed, silicon dangling bonds are the dominant defects in SiNx with x
